Expectations, Credibility and Disinflation

Why a shock becomes persistent

Inflation persistence is a wage-price process rather than a price-level fact. An energy spike arrives as a one-off cost and raises the price level, which is not inflation in the persistent sense. It becomes persistent when workers seek compensation for it and firms grant it and then reprice for their own costs, so the one-off becomes a rise in the rate at which costs grow. That is why the level of inflation expectations in the labour market, and specifically the compensation series and the wage agreements, is the single most informative thing to read about whether a disinflation is real. The anchor is the long-run expectation, and it is remarkably stable in credible regimes: households will report elevated one-year expectations while three-year expectations barely move, and that gap is exactly the measure of anchor credibility. When short and long expectations start to move together, the bank has lost the argument, and the arithmetic of restoring it changes — because expectations are in the wage-setting process, the only way to break them is to produce enough unemployment that wage setters no longer believe the high inflation will persist. That is the sacrifice ratio, and it is measured in output rather than in resolve. There is a distributional dimension worth holding alongside the arithmetic. Inflation that is expected and incorporated into contracts is a transfer; inflation that arrives as a surprise changes real wages and the real value of debt without any negotiation, which is why a surprise inflation episode leaves a permanent mark on the wealth distribution and why the politics of a disinflation are difficult regardless of the economics. Both facts belong in the account: the arithmetic says credibility is cheap, and the politics say that producing a recession to restore it is expensive in a way that does not appear in an output gap. The same 5% inflation, three exits — Supply reversal, expectations intact: inflation falls with no rise in unemployment — the cheap path ← · Demand slowdown: inflation falls, unemployment rises, and the cost is the output gap · Expectations unanchored: the bank must produce enough slack to be believed — the expensive path · What to watch: long-run expectations, compensation and wage agreements, not the headline A central bank cannot cut the rate that affects mortgages and wages by fiat; it moves the overnight rate and credibility does the rest. When credibility is weak, the same decision has less effect — which is why the first stage of a disinflation often requires a rate path higher than the arithmetic of the gaps would suggest.

Judging whether a disinflation is real

Four observations, read together. First, the composition: is the fall spread across core goods and core services, or concentrated in the volatile categories? Diffusion is the sign of a genuine turn. Second, the labour market: are wages decelerating without a rise in unemployment, which is the benign version, or only as unemployment rises, which is the expensive version? Third, expectations: have long-run expectations moved at all, and have they moved in the direction of the target? And fourth, the market: what has break-even inflation done, since it is the traded expectation rather than a survey response and it re-prices every day. The order matters because the four can disagree, and the disagreement is the information. A falling headline with stable long-run expectations and decelerating wages is a benign disinflation. A falling headline with rising long-run expectations is a pause rather than a turn. Rising unemployment with sticky wages is the worst configuration, because the bank is paying the cost without collecting the benefit — and it is the configuration most likely to produce a policy error, either a premature easing that restarts inflation or an over-tightening that turns a slowdown into something worse. For an investor, the practical implication is that the level of the policy rate is less informative than the credibility behind it. A regime with anchored expectations can support higher equity multiples at the same nominal rate, because the discount rate is stable in real terms; a regime where the anchor is wobbling requires a higher real rate to deliver the same outcome, which compresses multiples for reasons that have nothing to do with earnings. That is the channel through which a loss of credibility shows up in a portfolio before it shows up in a recession, and it is why the expectations series deserves a place in a monthly routine. • Diffusion across core categories is the sign of a real turn. • Wages falling without unemployment rising is benign; wages falling only with rising unemployment is expensive. • Long-run expectations moving is the warning; short-run expectations always move. • Break-even inflation is the market’s expectation, and it re-prices daily. The historical reference points are the two great disinflations: the early 1980s, where credibility had to be rebuilt with a deep recession, and the mid-2020s, where expectations held and the cost was smaller than most forecasts. The difference between them is not the size of the shock but the state of the anchor.

How expectations are measured — and what the measures miss

Expectations are unobservable, so every number used to describe them is a proxy with a known defect. The **survey** measures ask households or professional forecasters what they expect: the University of Michigan survey is the household series most quoted, and a central bank survey of consumer expectations tracks the same population with a longer horizon. Surveys have the advantage of directly asking, and two disadvantages: households are not always acting on the prices they are asked about, and the responses are sensitive to the price of petrol on the day the survey is fielded, so a single month of a survey can be a story about the pump rather than about expectations. The **market** measure comes from the difference in yield between a nominal bond and an inflation-linked bond of the same maturity — a breakeven. It is what someone is willing to pay to hedge inflation, which is not the same thing as their expectation of it. The yield difference contains a risk premium for holding the less liquid instrument, a premium that moves with demand, and a mismatch in the index the bond references against the index you care about. So a breakeven that rises can mean expected inflation rose, or that the inflation risk premium rose because people want insurance, and the two have very different implications for policy. Reading the market measure properly requires a model to pull the two apart, and the answer depends on the model. The reason this matters is that the single most important question about inflation expectations — are they anchored — is answered by the *long-horizon* measures and not the short ones. Short-horizon expectations rise in every energy shock and mean little; a long-run series that stays near target through a period of high realised inflation is the evidence that credibility held. In the 2022–2026 disinflation the long-run measures barely moved while short-run inflation ran near 9%, which is exactly the pattern the credibility argument predicts, and it is why the disinflation cost less output than a purely backward-looking relationship would have implied. When you check whether expectations are anchoring today, look at the five-year-ahead and ten-year-ahead figures — survey and market, together — and treat the one-year numbers as weather. • Surveys ask directly, but households react to today’s prices and may not act on the answers they give. • Breakevens come from the market, and include an inflation risk premium and a liquidity effect — they are not a pure expectation. • Short-horizon measures move with every energy shock and carry little information. • Anchoring is judged on the long horizon: it is the series that stayed still while realised inflation ran that made the disinflation cheap. Triangulate rather than choose: a long-run survey expectation, a long-run market breakeven, and the central bank’s own projections. When all three move together the signal is real; when one of them moves alone, the first question is what is wrong with that measure.

What a breakeven actually contains

The traded expectation is built by subtracting one bond from another: a nominal Treasury yield minus the yield on the inflation-linked bond of the same maturity gives a **breakeven inflation rate**, the compensation the market demands to hold nominal rather than real duration. The convenience of the measure is that it updates every second, unlike a survey. The trap is that it is not a pure expectation, and reading it as one is the most common error in this part of the subject. A breakeven is expected inflation **plus** an inflation risk premium, **minus** a liquidity premium for the inflation-linked bond, and the two adjustments can push it in the same direction or in opposite ones. Inflation risk is asymmetric for a nominal holder — surprises tend to be unpleasant and correlated with bad times for a portfolio — so investors historically demanded a premium, which means the breakeven sat above expected inflation. Inflation-linked bonds, meanwhile, are less liquid than the nominal market, and their holders accept a yield disadvantage, which pushes the breakeven below expected inflation. The net sign is an empirical question that has changed with the regime: a breakeven of two and a half percent may mean the market expects two and a half, or it may mean it expects two point three and is charging point two for inflation risk. Two devices separate the pieces. The first is the **forward breakeven**: instead of the whole maturity, compute the breakeven for the period from year five to year ten, which strips out the near-term path and the immediate policy cycle. The 5-year, 5-year forward has become the standard policy-watchers’ measure because it is far enough out to be free of the current energy price and close enough to be about the regime the committee is trying to hold. The second is to compare the market measure with a **survey** — professional forecasters, or the long-run expectations in a household survey. When the traded breakeven moves and the survey does not, the move is being driven by the premium or by liquidity rather than by a change in what people expect, which is the same distinction as the term premium against expectations in the curve lesson, applied to inflation instead of to rates. When they move together, the market has repriced the regime itself. The reason this matters for the disinflation question is that the anchoring evidence in this lesson — compensation, surveys and breakevens moving together — is only as strong as the weakest of the three, and the breakeven is the one that can be noisy for reasons that have nothing to do with expectations. A month in which the inflation-linked market seizes up, or in which the Treasury’s issuance of linkers changes the float, will move the breakeven without any change in what the market believes about prices. So the practical discipline is to read the breakeven against its own recent range, to prefer the five-year forward to the spot measure, to cross-check with a survey before drawing a conclusion about credibility, and to name the decomposition whenever the number is used as evidence — because “the market expects inflation of two point six percent” is a stronger claim than the data supports on its own. • A breakeven is expected inflation plus an inflation risk premium minus an inflation-liquidity premium. • The 5-year, 5-year forward strips the near-term path and is the standard regime measure. • When the breakeven moves and the survey does not, the move is premium or liquidity, not expectations. • Anchor evidence is only as strong as its weakest input, and the traded measure is the noisy one. • Read the breakeven against its own range and name the decomposition before treating it as a belief. The link back to the curve lesson is exact: the same curve can steepen because expectations rose or because a premium rose, and this is the inflation-sensitive version of that problem. The two decompositions share nothing but the discipline — a single traded number is a price that bundles a forecast with the compensation for being wrong about the forecast.

Sources

• Expectations, the Phillips curve and the sacrifice ratioFederal Reserve education materials; standard Phillips-curve literature
• The cost of disinflation under differing credibilityVolcker-era evidence; Sargent (1982) on the ends of four big inflations
• Anchoring of inflation expectations and the evidence for itFederal Reserve Bank of New York Survey of Consumer Expectations; market break-even inflation